I’m not sure anyone is actually convinced by long Facebook posts, but I enjoy digging into the data and unlike many, I try hard to share my thought processes and assumptions. I’ll apologize ahead of time since this is likely to be long, but I hope this ends up being one of the more definitive deep dives into the subject of COVID-19.

I see a lot of people saying that this virus isn’t a threat or is clearly much less of a threat than originally expected. This is typically used to justify moving away from the current partial shutdown very quickly. I have two problems with this argument.

1. I believe the data is clear that the virus is a deadly threat.

2. Even while the virus is a threat, we can manage the outbreak effectively without the heavy-handed government measures.

I’ll get to both topics eventually, but I’m going to start with the actual threat of the virus.

I started writing about COVID-19 in detail on March 27th. Since then, I’ve done original research into the publicly available data, particularly in Ohio, read various papers summarizing relevant studies and learned much from my collaboration with my brilliant friend Dr. Robert Bulas. In spite of all of the learning we have done since then, my conclusions haven’t changed much even as I know have much stronger evidence to support them than before.

When we talk about the threat of COVID-19, we have to consider how widely the virus can/will spread and the infection fatality rate (IFR). For a new virus, neither of these variables are well known up front. Originally, the expectation was that COVID-19 is a novel virus. This means that it is newly circulating within humans and it was presumed that we would have little built in immunity. In this case, the virus will typically circulate until enough of the population is infected to build herd immunity typically somewhere north of 50% of the total population. There were also some early estimates of the infection fatality rate of around 1%.

Let’s examine those assumptions based on what we know today. In Ohio, we have 16,601 lab confirmed cases and 702 probable cases out of about 11.9 million Ohioans as of 4/29. That amounts to 0.15% of our population. Of those 17,303 cases, 937 have passed so far for a case fatality rate of 5.4%. Notice that I changed terms. The case fatality rate is the measure of total deaths / total cases. It’s based on who we’ve actually tested or clinically diagnosed as having the disease. The infection fatality rate is based on the actual number of infections. Unless we test and retest all 11.9 million Ohioans, we’ll never know this number for sure. Instead, we have to find a way to estimate the actual number of cases. On the contrary, Ohio has done a pretty good job of testing those who get very sick. We can be reasonably confident the 937 deaths represent a fair number of actual deaths for the state.

We’ve got a couple of methods that let us infer the actual number of infections. The first option is to use antibody testing. In this method, we test a group to determine if their blood contains antibodies to the virus and compare the results to those known to have the virus. If you are currently or previously infected, you will likely show positive on this test. I’ve previously shared several of these studies that have been done in various parts of the country and they show that the actual number of infected people is between 20 and 80 times the measured number of cases. We don’t have this data for Ohio, but to give you an example of this impact, 20x cases means our CFR of 5.4% becomes an IFR of 0.28%. The black plague essentially becomes something slightly more deadly than the flu.

The other way we can estimate the IFR is to use closed populations where we have mass testing. We have two of these scenarios that I’ve been able to find – the Diamond Princess cruise ship, and Ohio’s prison system. In both scenarios, nearly everyone in these environments was tested, so we know the actual number of infections and we can calculate the true IFR. For the Diamond Princess, 17% of the ship was infected in spite of a somewhat quick lockdown process and the IFR was 1.1%. In Ohio, a facility in Pickaway County and another in Marion County ended up having 80% of inmates become infected. 3461 total inmates are currently infected of which 23 have passed away for a current IFR of 0.66%. Many of these cases are less than two weeks old, so it’s likely we’ll see some additional deaths to push that number higher over the next several days.

It’s fair to say that inmates are probably not the most healthy population overall and that neither the cruise ship nor prisons have an age distribution identical to the rest of the population. That being said, they aren’t all that far off. The cruise ship included many young staff members that offset that the older passengers and the prisons I’m studying had a surprisingly (to me anyways) older population. To the extent that we adjust for the age distributions of these groups, we are probably understating the IFR to a small amount.

So to summarize, we have various estimates of the true IFR ranging from 0.07% at the low end (similar to flu) to 1.1% at the high end (11x deadlier) with the stronger evidence (actual measured cases are better than extrapolations from antibody testing) implying that we’re at the higher end of that range. We also see that the virus is highly contagious and can infect as much as 80% of a population if it spreads quickly in close quarters. Given that the flu is estimated to infect a max of 14% of USA population in a given year, my conclusion is that COVID-19 is a virus that is a significant threat especially considering rapid spread runs the risk of overwhelming hospital systems and driving a much higher IFR than we would see otherwise.

A few other thoughts before I wrap this up:

1. The IFR is not a static number. The IFR can change over time depending on how we evolve treatments. If they get more effective, the IFR will decline. Similarly, it’s possible that the people who get the virus first are more vulnerable and will have a higher death rate than those that get it later. It’s also possible that low level exposure over time builds up some level of immunity that is helpful to people who come down with the virus later. I haven’t seen evidence yes or no that this is the case for COVID-19, but I think it’s likely to be true.

2. A known IFR implies a true number of cases. If we assume the actual IFR is 1%, we can then assume that there are 100 actual cases for each death. In Ohio, that would mean that we really have 93,700 cases, nearly 5.5x the reported number. If the actual IFR is 0.5%, we’re now at 187,400 cases. Even with the latter number, that implies that 1.6% of Ohio’s population has been infected and that we are very early in the progress of the outbreak towards developing herd immunity.

3. Ohio Deaths at various levels of IFR and infected population:

Infection Fatality Rate Infected Population Death Count
1.0% 50% 59,500
1.0% 15% 17,850
0.5% 50% 29,750
0.5% 15% 8,925
0.1% 50% 5,950
0.1% 15% 1,875

Part 2 will be on the policy implications of the above. Yes, I will address the varying risks to different age levels, the Ohio mask controversy and more.